From simplified text to knowledge representation using controlled natural language

Abstract

Knowledge based systems provide means to store data and perform reasoning on top of it. Controlled Natural Language (CNL) is considered as an engineered subset of natural language. CNLs aim to abstract the complexity level of natural languages and reduce/abolish the characterizing ambiguity of these languages. Two major types of CNLs exist. The one aiming to improve human-human communication is called Simplified Language or human CNL. The second type is a machinereadable CNL, which is formal language used mainly as an Abstract Knowledge Representation (AKR) Language. In this paper, we present an approach to translate the Medline summaries of diseases presented in a Simplified English to a machine readable CNL. This approach will result in a formal Knowledge Base enabling machine processing, querying and reasoning while avoiding the usual main obstacle which is natural language ambiguity. We tested the approach on a sample corpus of the Anemia disease, and presented the concluded hypothesis.